Skip to main content
. Author manuscript; available in PMC: 2023 Jan 3.
Published in final edited form as: Med Image Anal. 2022 May 25;80:102484. doi: 10.1016/j.media.2022.102484

Table 2.

Evaluation of the components of the proposed automated DSVR reconstruction pipeline on 5 severe rotation and translation motion datasets with respect to the % proportion of rejected slices and 2D NCC between simulated and original slices in all stacks and only the excluded stack. The investigated scenarios include: (i) Steps I+IV; (ii) Steps I+II+IV; (iii) full pipeline: Steps I+II+III+IV. The results are statistically significant with p < .001 for comparison between (i) vs. (ii) and (iii) scenarios and with p < .01 for (ii) vs. (iii) comparison. Step I is thorax localisation, Step II is reorientation to the atlas, Step III is stack selection and average template generation and Step IV is DSVR reconstruction.

Scenario % of rejected all NCC excl. NCC
I+IV 64.7 ± 5.3% 0.579 ± 0.058 0.550 ± 0.218
I+II+IV 51.6 ± 8.0% 0.728 ± 0.090 0.705 ± 0.105
I+II+III+IV 37.8 ± 9.1% 0.834 ± 0.028 0.809 ± 0.058